{"url":"/sota/machine-translation-on-iwslt2014-german","task":{"name":"Machine Translation","url":"/task/machine-translation","note":null},"dataset":{"name":"IWSLT2014 German-English","url":null},"category":"Natural Language Processing","categories":["Natural Language Processing"],"category_note":null,"description":"**Machine translation** is the task of translating a sentence in a source language to a different target language.\r\n\r\nApproaches for machine translation can range from rule-based to statistical to neural-based. More recently, encoder-decoder attention-based architectures like BERT have attained major improvements in machine translation. \r\n\r\nOne of the most popular datasets used to benchmark machine translation systems is the WMT family of datasets. Some of the most commonly used evaluation metrics for machine translation systems include BLEU, METEOR, NIST, and others. \r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [Google seq2seq](https://github.com/google/seq2seq) )</span>","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["BLEU score","Number of Params"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"BLEU score":"higher","Number of Params":"lower"}},"counts":{"rows":34,"rows_with_code":29,"rows_with_paper_page":34,"rows_dated":34,"rows_using_additional_data":2},"rows":[{"rank_in_archive_order":1,"model":"PiNMT","metrics":{"BLEU score":"40.43"},"uses_additional_data":false,"paper_date":"2023-10-30","paper":"/paper/integrating-pre-trained-language-model-into","paper_url":"https://arxiv.org/abs/2310.19680v4","paper_title":"Integrating Pre-trained Language Model into Neural Machine Translation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"BiBERT","metrics":{"BLEU score":"38.61","Number of Params":"73.8M"},"uses_additional_data":false,"paper_date":"2021-09-09","paper":"/paper/bert-mbert-or-bibert-a-study-on","paper_url":"https://arxiv.org/abs/2109.04588v1","paper_title":"BERT, mBERT, or BiBERT? A Study on Contextualized Embeddings for Neural Machine Translation","code":"https://github.com/fe1ixxu/BiBERT","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":1}},{"rank_in_archive_order":3,"model":"Bi-SimCut","metrics":{"BLEU score":"38.37"},"uses_additional_data":false,"paper_date":"2022-06-06","paper":"/paper/bi-simcut-a-simple-strategy-for-boosting-1","paper_url":"https://arxiv.org/abs/2206.02368v2","paper_title":"Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation","code":"https://github.com/gpengzhi/Bi-SimCut","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"Cutoff + Relaxed Attention + LM","metrics":{"BLEU score":"37.96","Number of Params":"24.1M"},"uses_additional_data":true,"paper_date":"2022-09-20","paper":"/paper/relaxed-attention-for-transformer-models","paper_url":"https://arxiv.org/abs/2209.09735v1","paper_title":"Relaxed Attention for Transformer Models","code":"https://github.com/Oguzhanercan/Vision-Transformers","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"DRDA","metrics":{"BLEU score":"37.95"},"uses_additional_data":false,"paper_date":"2024-06-04","paper":"/paper/deterministic-reversible-data-augmentation","paper_url":"https://arxiv.org/abs/2406.02517v1","paper_title":"Deterministic Reversible Data Augmentation for Neural Machine Translation","code":"https://github.com/BITHLP/DRDA","n_code_links":1,"syntology":null},{"rank_in_archive_order":6,"model":"Transformer + R-Drop + Cutoff","metrics":{"BLEU score":"37.90"},"uses_additional_data":false,"paper_date":"2021-06-28","paper":"/paper/r-drop-regularized-dropout-for-neural","paper_url":"https://arxiv.org/abs/2106.14448v2","paper_title":"R-Drop: Regularized Dropout for Neural Networks","code":"https://github.com/dropreg/R-Drop","n_code_links":8,"syntology":{"n_ran":2,"n_unverified":4,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":7,"model":"SimCut","metrics":{"BLEU score":"37.81"},"uses_additional_data":false,"paper_date":"2022-06-06","paper":"/paper/bi-simcut-a-simple-strategy-for-boosting-1","paper_url":"https://arxiv.org/abs/2206.02368v2","paper_title":"Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation","code":"https://github.com/gpengzhi/Bi-SimCut","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"Cutoff+Knee","metrics":{"BLEU score":"37.78"},"uses_additional_data":false,"paper_date":"2020-03-09","paper":"/paper/wide-minima-density-hypothesis-and-the","paper_url":"https://arxiv.org/abs/2003.03977v5","paper_title":"Wide-minima Density Hypothesis and the Explore-Exploit Learning Rate Schedule","code":"https://github.com/HA-Transformer/HA-Transformer","n_code_links":5,"syntology":null},{"rank_in_archive_order":9,"model":"Cutoff","metrics":{"BLEU score":"37.6"},"uses_additional_data":false,"paper_date":"2020-09-29","paper":"/paper/a-simple-but-tough-to-beat-data-augmentation","paper_url":"https://arxiv.org/abs/2009.13818v2","paper_title":"A Simple but Tough-to-Beat Data Augmentation Approach for Natural Language Understanding and Generation","code":"https://github.com/dinghanshen/Cutoff","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":10,"model":"CipherDAug","metrics":{"BLEU score":"37.53"},"uses_additional_data":false,"paper_date":"2022-04-01","paper":"/paper/cipherdaug-ciphertext-based-data-augmentation","paper_url":"https://arxiv.org/abs/2204.00665v1","paper_title":"CipherDAug: Ciphertext based Data Augmentation for Neural Machine Translation","code":"https://github.com/protonish/cipherdaug-nmt","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"Transformer + R-Drop","metrics":{"BLEU score":"37.25"},"uses_additional_data":false,"paper_date":"2021-06-28","paper":"/paper/r-drop-regularized-dropout-for-neural","paper_url":"https://arxiv.org/abs/2106.14448v2","paper_title":"R-Drop: Regularized Dropout for Neural Networks","code":"https://github.com/dropreg/R-Drop","n_code_links":8,"syntology":{"n_ran":2,"n_unverified":4,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":12,"model":"Data Diversification","metrics":{"BLEU score":"37.2"},"uses_additional_data":false,"paper_date":"2019-11-05","paper":"/paper/data-diversification-an-elegant-strategy-for","paper_url":"https://arxiv.org/abs/1911.01986v4","paper_title":"Data Diversification: A Simple Strategy For Neural Machine Translation","code":"https://github.com/nxphi47/data_diversification","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":1,"n_samples":4,"n_pointer_only_licence":4}},{"rank_in_archive_order":13,"model":"UniDrop","metrics":{"BLEU score":"36.88"},"uses_additional_data":false,"paper_date":"2021-04-11","paper":"/paper/unidrop-a-simple-yet-effective-technique-to","paper_url":"https://arxiv.org/abs/2104.04946v1","paper_title":"UniDrop: A Simple yet Effective Technique to Improve Transformer without Extra Cost","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":14,"model":"MixedRepresentations","metrics":{"BLEU score":"36.41"},"uses_additional_data":false,"paper_date":"2020-07-11","paper":"/paper/sequence-generation-with-mixed","paper_url":"https://icml.cc/Conferences/2020/ScheduleMultitrack?event=6391","paper_title":"Sequence Generation with Mixed Representations","code":"https://github.com/apeterswu/fairseq_mix","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"Mask Attention Network (small)","metrics":{"BLEU score":"36.3","Number of Params":"37M"},"uses_additional_data":false,"paper_date":"2021-03-25","paper":"/paper/mask-attention-networks-rethinking-and","paper_url":"https://arxiv.org/abs/2103.13597v1","paper_title":"Mask Attention Networks: Rethinking and Strengthen Transformer","code":"https://github.com/libertfan/man","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":3,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"MUSE(Parallel Multi-scale Attention)","metrics":{"BLEU score":"36.3"},"uses_additional_data":false,"paper_date":"2019-11-17","paper":"/paper/muse-parallel-multi-scale-attention-for","paper_url":"https://arxiv.org/abs/1911.09483v1","paper_title":"MUSE: Parallel Multi-Scale Attention for Sequence to Sequence Learning","code":"https://github.com/lancopku/MUSE","n_code_links":3,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":17,"model":"Transformer+Rep(Sim)+WDrop","metrics":{"BLEU score":"36.22","Number of Params":"37M"},"uses_additional_data":false,"paper_date":"2021-04-05","paper":"/paper/rethinking-perturbations-in-encoder-decoders","paper_url":"https://arxiv.org/abs/2104.01853v1","paper_title":"Rethinking Perturbations in Encoder-Decoders for Fast Training","code":"https://github.com/takase/rethink_perturbations","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":18,"model":"MAT","metrics":{"BLEU score":"36.22"},"uses_additional_data":false,"paper_date":"2020-06-18","paper":"/paper/multi-branch-attentive-transformer","paper_url":"https://arxiv.org/abs/2006.10270v2","paper_title":"Multi-branch Attentive Transformer","code":"https://github.com/HA-Transformer/HA-Transformer","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":19,"model":"TransformerBase + AutoDropout","metrics":{"BLEU score":"35.8"},"uses_additional_data":false,"paper_date":"2021-01-05","paper":"/paper/autodropout-learning-dropout-patterns-to","paper_url":"https://arxiv.org/abs/2101.01761v1","paper_title":"AutoDropout: Learning Dropout Patterns to Regularize Deep Networks","code":"https://github.com/google-research/google-research","n_code_links":1,"syntology":null},{"rank_in_archive_order":20,"model":"Local Joint Self-attention","metrics":{"BLEU score":"35.7"},"uses_additional_data":false,"paper_date":"2019-05-16","paper":"/paper/190506596","paper_url":"https://arxiv.org/abs/1905.06596v1","paper_title":"Joint Source-Target Self Attention with Locality Constraints","code":"https://github.com/jarfo/joint","n_code_links":2,"syntology":null},{"rank_in_archive_order":21,"model":"TaLK Convolutions","metrics":{"BLEU score":"35.5"},"uses_additional_data":false,"paper_date":"2020-02-08","paper":"/paper/time-aware-large-kernel-convolutions","paper_url":"https://arxiv.org/abs/2002.03184v2","paper_title":"Time-aware Large Kernel Convolutions","code":"https://github.com/lioutasb/TaLKConvolutions","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":2,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":22,"model":"ImitKD + Full","metrics":{"BLEU score":"35.4"},"uses_additional_data":false,"paper_date":"2020-09-15","paper":"/paper/autoregressive-knowledge-distillation-through","paper_url":"https://arxiv.org/abs/2009.07253v2","paper_title":"Autoregressive Knowledge Distillation through Imitation Learning","code":"https://github.com/asappresearch/imitkd","n_code_links":2,"syntology":null},{"rank_in_archive_order":23,"model":"DeLighT","metrics":{"BLEU score":"35.3"},"uses_additional_data":false,"paper_date":"2020-08-03","paper":"/paper/delight-very-deep-and-light-weight","paper_url":"https://arxiv.org/abs/2008.00623v2","paper_title":"DeLighT: Deep and Light-weight Transformer","code":"https://github.com/sacmehta/delight","n_code_links":2,"syntology":{"n_ran":0,"n_unverified":3,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":24,"model":"DynamicConv","metrics":{"BLEU score":"35.2"},"uses_additional_data":false,"paper_date":"2019-01-29","paper":"/paper/pay-less-attention-with-lightweight-and","paper_url":"http://arxiv.org/abs/1901.10430v2","paper_title":"Pay Less Attention with Lightweight and Dynamic Convolutions","code":"https://github.com/pytorch/fairseq","n_code_links":3,"syntology":null},{"rank_in_archive_order":25,"model":"Transformer","metrics":{"BLEU score":"35.1385"},"uses_additional_data":false,"paper_date":"2022-05-15","paper":"/paper/guidelines-for-the-regularization-of-gammas","paper_url":"https://arxiv.org/abs/2205.07260v1","paper_title":"Guidelines for the Regularization of Gammas in Batch Normalization for Deep Residual Networks","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":26,"model":"LightConv","metrics":{"BLEU score":"34.8"},"uses_additional_data":false,"paper_date":"2019-01-29","paper":"/paper/pay-less-attention-with-lightweight-and","paper_url":"http://arxiv.org/abs/1901.10430v2","paper_title":"Pay Less Attention with Lightweight and Dynamic Convolutions","code":"https://github.com/pytorch/fairseq","n_code_links":3,"syntology":null},{"rank_in_archive_order":27,"model":"Transformer","metrics":{"BLEU score":"34.44"},"uses_additional_data":false,"paper_date":"2017-06-12","paper":"/paper/attention-is-all-you-need","paper_url":"https://arxiv.org/abs/1706.03762v7","paper_title":"Attention Is All You Need","code":"https://github.com/huggingface/transformers","n_code_links":595,"syntology":{"n_ran":600,"n_unverified":346,"n_samples":946,"n_pointer_only_licence":451}},{"rank_in_archive_order":28,"model":"Rfa-Gate-arccos","metrics":{"BLEU score":"34.4"},"uses_additional_data":false,"paper_date":"2021-03-03","paper":"/paper/random-feature-attention-1","paper_url":"https://arxiv.org/abs/2103.02143v2","paper_title":"Random Feature Attention","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":29,"model":"Variational Attention","metrics":{"BLEU score":"33.1"},"uses_additional_data":false,"paper_date":"2018-07-10","paper":"/paper/latent-alignment-and-variational-attention","paper_url":"http://arxiv.org/abs/1807.03756v2","paper_title":"Latent Alignment and Variational Attention","code":"https://github.com/harvardnlp/var-attn","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":1,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":30,"model":"Minimum Risk Training [Edunov2017]","metrics":{"BLEU score":"32.84"},"uses_additional_data":false,"paper_date":"2017-11-14","paper":"/paper/classical-structured-prediction-losses-for","paper_url":"http://arxiv.org/abs/1711.04956v5","paper_title":"Classical Structured Prediction Losses for Sequence to Sequence Learning","code":"https://github.com/pytorch/fairseq","n_code_links":1,"syntology":null},{"rank_in_archive_order":31,"model":"CNAT","metrics":{"BLEU score":"31.15"},"uses_additional_data":false,"paper_date":"2021-03-21","paper":"/paper/non-autoregressive-translation-by-learning","paper_url":"https://arxiv.org/abs/2103.11405v1","paper_title":"Non-Autoregressive Translation by Learning Target Categorical Codes","code":"https://github.com/baoy-nlp/CNAT","n_code_links":1,"syntology":null},{"rank_in_archive_order":32,"model":"Neural PBMT + LM [Huang2018]","metrics":{"BLEU score":"30.08"},"uses_additional_data":false,"paper_date":"2017-06-17","paper":"/paper/towards-neural-phrase-based-machine","paper_url":"http://arxiv.org/abs/1706.05565v8","paper_title":"Towards Neural Phrase-based Machine Translation","code":"https://github.com/posenhuang/NPMT","n_code_links":4,"syntology":null},{"rank_in_archive_order":33,"model":"Back-Translation Finetuning","metrics":{"BLEU score":"28.83"},"uses_additional_data":true,"paper_date":"2019-12-22","paper":"/paper/tag-less-back-translation","paper_url":"https://arxiv.org/abs/1912.10514v3","paper_title":"Tag-less Back-Translation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":34,"model":"Actor-Critic [Bahdanau2017]","metrics":{"BLEU score":"28.53"},"uses_additional_data":false,"paper_date":"2016-07-24","paper":"/paper/an-actor-critic-algorithm-for-sequence","paper_url":"http://arxiv.org/abs/1607.07086v3","paper_title":"An Actor-Critic Algorithm for Sequence Prediction","code":"https://github.com/joeynmt/joeynmt","n_code_links":3,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":0}}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,264 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9581,"papers_checked":6264,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":3316},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":14,"rows_with_any_sample_ran":10,"distinct_papers_with_graph_line":13,"distinct_papers_with_any_sample_ran":9,"samples_over_distinct_papers":{"n_ran":617,"n_unverified":363,"n_samples":980,"n_pointer_only_licence":462,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":619,"n_unverified":367,"n_samples":986,"n_pointer_only_licence":462,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}